When I was enlisted to instruct a Design of Experiments course for a bioprocessing meeting in 2009, I first started to piece together slides for a “typical” approach. The more I got into the meat of the subject, however, the more I realized that biotech is not the same as chemical or even pharmaceutical manufacturing. There is one major difference: the “reaction mechanism” is not a single, purely chemical reaction; it is, in most cases, based on the hard work of millions of “wee little beasties.” What does this mean in practical terms?
One immediate consequence is that we do not have much freedom in the parameters (such as temperature) that we can manipulate in a DoE. (Unfriendly conditions tend to kill our little friends.) The catalysts we use in a biotech reaction are often themselves results of bio-reactions and are often poorly characterized. By comparison, the typical platinum or palladium catalysts, used in chemical reactions, with a measurable/controllable surface area, are a heck of a lot easier to control than, say, sludge from a brewery process.
Even the tests used to control a bioprocess are often seen as “insurmountable opportunities.” While chemical and pharmaceutical analytical chemists have increased the speed and precision of their analytical tools, the biotech people are still using, in essence, tests from deep in the last century even as they upgrade their fermentation hardware. (It is interesting to note that the pharma industry uses the most modern analytical hardware, yet produces the tablets and capsules much like “grand-daddy did” in 1960.) Traditional hours-long or days-long biological tests do not lend themselves to either PAT or analyzing the large number of runs required for a proper DoE. It would appear that, while pharma can make its analytical tests into “process” tests by miniaturization and making the hardware wireless, it is not so easy to speed up biotech analysis methodology.
Indeed, the methodologies common to pharma actually can be adopted by biopharma folks. This is where the differences between pharma and biopharma are highlighted. However, adaption of these methods to biotech is not that simple. In a tablet (blend, granulation, and so forth), we are simply looking for a dry chemical or two, often in the percentage range, in a fixed, relatively simple matrix. In a bioprocess, the water alone is a problem for many spectrometric approaches. Add to that the complexity of the “soup” (the matrix changes throughout the process) and that the product strongly resembles its components, requires you to do some serious correlation math between any spectrum or other data output and conventional method results. (Of course these correlations are the meat and potatoes of multivariate analyses, the basis of DoE.)
In our well-known pharma production model, we have strongly invested in “traditional” methodology for producing the final product. Any actual chemistry is done in the API stage, now often outsourced. The final drug “product” is made by, if the truth be told, a complex mixing process. Considering the complexity of blending disparate powders, do we hire physical chemists, engineers, and materials scientists to design these delivery systems? No, we hire pharmacists.
Now, to borrow a phrase, some of my best friends are pharmacists. It wasn’t all that many years ago that one got his/her pharmacy degree in two years. And, despite many upgrades to the curriculum, pharmacists still spend most of their time learning about the physiological properties of the drugs themselves rather than the physics needed to make a good product mixture. As an unintended consequence, the people who make the product must depend on another set of scientists to tell them how well the product was made: the Quality Control analysts.
On the other hand, scientists who design and execute the production of biopharmaceutical products are, in fact, the same (or with the same training) people who analyze them for potency, purity, etc. There is less of a ”vested interest” in keeping a stable of scientists who do nothing other than see if the product was made properly. In other words, the assay is just “another service” that the busy biotech people have to provide in addition to all their other duties. They are, then, more open to easier ways to do things.
So, what does all this mean for the rapid implementation of PAT into each industry? Well, we have fewer technical difficulties in traditional drug manufacturing, but confront more inertia because it may be viewed by some departments as intruding on their “turf.” The biopharma folks (with whom I’ve worked with so far) welcome anything that lightens their work load and speeds up incoming information, helping to actually control their product. Therefore, I see biopharma as the easier path for PAT taking root in the near future.
